Silly question about ECC

hello, as from title, I have a silly question about the ECC memory.

can I have it only on the storage machine? or to avoid any problem, all the machines must have it?

I don’t know how exaclty explain it in english but let suppose that I have this configuration:

  • PCs (no ECC) that runs vm and Docker or k3s
  • server (ECC) that run Truenas

If a bit flip occurs on the PC side, will the data sent to truenas fixed by the server ram before put it on the drive or I still have the risk of a corruption and loose all the data in the pool?
Or I am totally out of road and this is not how it works?

Because since I am planning to rebuild the homelab (to reduce the power consumption) the idea was to use some nuc or similar as worker and leave a small server with ECC for the storage.

One thing I would point out is ECC only protects you while the data is being processed or resides in memory. Once it sits on a storage device, the file system protects your data, not necessarily ECC memory.

If you flip a bit on the PC side, assuming it is new data and not data that already resides on TrueNAS, then TrueNAS will not be able to fix it.

Personally for my homelab, there is nothing about my services or my data that is so critical that I need ECC memory. I just run without it.

So, if I correctly understand the “damage” will only hit the new data and not the entire pool, am i right?

Me too but I would avoid to rescan all the house documents, photos, re-rip all the cd/bluray and re-download all the “linux distros” I got in the years.

Thats why I want to protect the nas as possible (ecc, external backups, etc)

So, if I correctly understand the “damage” will only hit the new data and not the entire pool, am i right?

I believe that is correct. There was a lot of fuss over the “ZFS scrub of death” but I think that has been pretty well dis-proven at this point. If data is sitting on a hard disk, and not being moved into or out of system memory for some reason, I fail to see any way that ECC system memory would have any effect on it. Hard disks have their own ECC built into them.

As far as re-ripping CDs, etc., that’s what a good back up strategy is for. RAID is not a backup, and system ECC only touches your data while it is being processed in some way. A good backup strategy has two copies on site and a third copy off site. I put my offsite copy into the cloud. A good cloud provider will store your data in more than one location, and their systems have all the usual enterprise protections.

1 Like